One of the most well-known cognitive biases is confirmation bias, wherein selection and evaluation of information that would confirm a focal hypothesis is given priority—or even exclusivity. Some researchers suggested that there was a strong confirmation bias in visual search, that attention is preferentially deployed to stimuli matching a target template, even when alternate strategies would reduce the number of searches necessary. However, visual search can be made more efficient by adopting the optimal strategy. There is growing consensus that reward plays an important role in the control of attention, even reward-associated stimuli can automatically capture attention. Therefore, goal-directed processes and stimulus-driven processes would produce optimal balance according to the changes of stimulus resources. Given the reaction time is indirect evidences for confirmation bias, in this article, we pursue the question of whether visual selection exhibits a confirmation bias by eye movements; specifically, when searching for a particular stimulus there will be a fixation for it. If there is not the fixation for target, but the response is correct, then there may be the change of visual search strategies. The present study used eye tracking to measure the effects of each type of attentional template representation on visual search processes. The experiment was a 2 (color match: match, mismatch) × 3 (red proportion: 25%, 50%, 75%) within-subject design. Participants are recruited to finish the visual search task, which the stimulus are composed of lowercase p’s, q’s, b’s, and d’s. Each letter has two colors (ren and blue). The participants’ task is to make a decision of the target letter’s color. During the search, dwell time, fixation and saccade amplitude were recorded. The results showed that there was a confirmation bias based on the findings of reaction time. But, when used the fixation as the index, the confirmation responses and optimal strategy responses showed some different changes. That is, with the difficulty of searching varies, participants showed the different performance: in the match condition, when the red letters’ proportion is 25% and 50%, participants reacted most quickly and the proportion of confirmation responses is the most. However, in the mismatch condition, the confirmation responses decreased and the optimal strategy responses increased significantly. Though the evidence to date supports the possibility that top-down visual selection mechanisms may automatically lead to confirmatory searching. By using the fixation as the index, we measured whether stimuli that could confirm the presence of a target are prioritized over those that could disconfirm the presence of a target. Our results provide support for the notion that the suitable strategies during visual search will be changed.The confirmation response based on template-target matching, and the strategy response will automatically switch according to the availability of resources to achieve the optimal performance. We take these results as suggesting that decision makers are assumed to have the intention to seek truth and make optimal decisions, but their decisions showed a remarkable flexibility on a trial-by-trial basis, the usefulness of such strategies depends on the type of template-target matching to be the most efficiently compared. Eye movements reflect and shape strategies for the visual search tasks automatically.